GHCi's claim of infixr 0 (->) is a lie

Currently, if you query the :info for (->) in GHCi, it will give you:

$ ghci
GHCi, version 8.4.3: http://www.haskell.org/ghc/  :? for help
Loaded GHCi configuration from /home/ryanglscott/.ghci
λ> :i (->)
data (->) (a :: TYPE q) (b :: TYPE r)   -- Defined in ‘GHC.Prim’
infixr 0 `(->)`
<instances elided>

This fixity information appears to be plain wrong, as the following program demonstrates:

{-# LANGUAGE TypeOperators #-}
module Bug where

import Data.Type.Equality

type (~>) = (->)
infixr 0 ~>

f :: (a ~> b -> c) :~: (a ~> (b -> c))
f = Refl

Since (~>) and (->) are both infixr 0, I would expect a ~> b -> c to associate as a ~> (b -> c), like the type signature for f wants to prove. However, GHC believes otherwise:

$ ghc Bug.hs                                                                                      
[1 of 1] Compiling Bug              ( Bug.hs, Bug.o )                                              
                                                                                                   
Bug.hs:10:5: error:                                                                                
    • Occurs check: cannot construct the infinite type: a ~ a ~> b                                 
      Expected type: ((a ~> b) -> c) :~: (a ~> (b -> c))
        Actual type: ((a ~> b) -> c) :~: ((a ~> b) -> c)
    • In the expression: Refl
      In an equation for ‘f’: f = Refl
    • Relevant bindings include
        f :: ((a ~> b) -> c) :~: (a ~> (b -> c)) (bound at Bug.hs:10:1)
   |
10 | f = Refl
   |     ^^^^

Reading the error message above, it appears that GHC gives (->) an even //lower// precedence than 0, since it associates a ~> b -> c as (a ~> b) -> c.

I'm not sure how to reconcile these two facts. There are at least a couple of options I can think of:

1. Claim (->) has a negative fixity.
2. Try to change GHC so that (->) really is infixr 0.

Trac metadata

Trac field	Value
Version	8.4.3
Type	Bug
TypeOfFailure	OtherFailure
Priority	normal
Resolution	Unresolved
Component	Compiler
Test case
Differential revisions
BlockedBy
Related
Blocking
CC
Operating system
Architecture

mentioned in issue #10056 (closed)

changed milestone to %8.6.1

changed weight to 5

added Tbug Trac import labels

I've been bitten by this before.

I'm worried about breakage with option (2), so I favor option (1), even if negative fixities are not available to users.

While we're in town, it should be infixr (-1) ->, without parentheses/backticks around the arrow.

Replying to [ticket:15235#comment:154620 goldfire]:

I'm worried about breakage with option (2), so I favor option (1), even if negative fixities are not available to users.

Sounds good to me.

While we're in town, it should be infixr (-1) ->, without parentheses/backticks around the arrow.

This was done separately in #15236 (closed).

Trac metadata

Trac field	Value
Related	- → #15236 (closed)

changed milestone to %8.8.1

These won't be addressed in GHC 8.6.

Changing GHCi's fixity information for (->) is a simple matter of applying this change:

diff --git a/compiler/basicTypes/BasicTypes.hs b/compiler/basicTypes/BasicTypes.hs
index 93010b7..e9d32f6 100644
--- a/compiler/basicTypes/BasicTypes.hs
+++ b/compiler/basicTypes/BasicTypes.hs
@@ -409,7 +409,7 @@ defaultFixity = Fixity NoSourceText maxPrecedence InfixL
 negateFixity, funTyFixity :: Fixity
 -- Wired-in fixities
 negateFixity = Fixity NoSourceText 6 InfixL  -- Fixity of unary negate
-funTyFixity  = Fixity NoSourceText 0 InfixR  -- Fixity of '->'
+funTyFixity  = Fixity NoSourceText (-1) InfixR  -- Fixity of '->'
 
 {-
 Consider

However, there's one more question we should answer before applying this change: do we want the displayed :info output to be this:

infixr -1 ->

Or this?

infixr (-1) ->

With only the above change, :info will display the former. If we want the latter, we'd have to make additional changes to the pretty-printer output for fixity information to add parentheses.

I don't think it matters much, so I'd be inclined to do whatever is least work.

Sounds good to me. This would be a good task for someone wanting to get involved with GHC development, so I'll mark this as a newcomer ticket. All you'd need to do is apply the change in ticket:15235#comment:155871, and include a comment explaining why this negative fixity is used.

added newcomer label

See https://github.com/ghc/ghc/pull/158 for a patch.

The patch in ticket:15235#comment:156377 has been superseded by D5199.

Trac metadata

Trac field	Value
Differential revisions	- → D5199

mentioned in commit 251e3424

closed

Trac metadata

Trac field	Value
Resolution	Unresolved → ResolvedFixed
Test case	- → ghci/scripts/T8535

added Pnormal label

mentioned in issue #17544 (closed)